Abstract: In the measurement of special terrain in mines, elevation information is crucial for accurately describing the terrain and is closely related to information such as mine slope and valley depth. Traditional 3D models relying on a single sensor can only present partial terrain features, making it difficult to accurately obtain elevation information of special terrain areas such as steep slopes and valleys, resulting in missing or incorrect information and significant measurement errors in these areas. Therefore, this study proposes a multi-sensor fusion technology for special terrain measurement in mines. Using unmanned aerial vehicles as carriers to carry multiple sensors to obtain surface images of mines, planning routes and setting parameters with the help of geographic information systems, using oblique photography to collect images from different angles, and obtaining point cloud data with airborne LiDAR to obtain special terrain measurement data of mines. Based on artificial intelligence algorithms for special terrain recognition, Adaboost algorithm is used to extract key information from images and point cloud data, identify special terrain, and provide a foundation for 3D modeling. Using Delaunay triangulation and texture mapping methods for 3D modeling of special terrain in mines. By using GNSS/IMU fusion coordinate transformation technology to obtain accurate elevation data, improve 3D model information, and ultimately construct a special terrain 3D model of the mine that combines precise terrain structure and real surface features. The results show that the error fluctuation curve of the studied method is generally close to the baseline (0 value), with small and stable fluctuation amplitude, high measurement accuracy, and stable performance, which can provide strong data support for mine planning, mining, and safety management.